[OE-core] [PATCH dylan, master] Fix meta-toolchain-qte SDK build for x11-less DISTRO

Thu Jun 6 16:28:08 UTC 2013

On Thu, Jun 6, 2013 at 12:46 PM, Richard Purdie <
richard.purdie at linuxfoundation.org> wrote:

> On Wed, 2013-06-05 at 16:58 -0400, Denys Dmytriyenko wrote:
> > On Wed, Jun 05, 2013 at 05:56:31PM -0300, Otavio Salvador wrote:
> > > On Wed, Jun 5, 2013 at 5:53 PM, Saul Wold <sgw at linux.intel.com> wrote:
> > >
> > > > On 06/05/2013 01:10 PM, Otavio Salvador wrote:
> > > >
> > > >>
> > > >>
> > > >>
> > > >> On Wed, Jun 5, 2013 at 5:07 PM, Denys Dmytriyenko <denis at denix.org
> > > >> <mailto:denis at denix.org>> wrote:
> > > >>
> > > >>     On Wed, Jun 05, 2013 at 02:43:07PM -0300, Otavio Salvador wrote:
> > > >>      > On Wed, Jun 5, 2013 at 2:23 PM, Nicolas Dechesne <
> > > >>      > nicolas.dechesne at linaro.org <mailto:nicolas.dechesne@**
> linaro.org<nicolas.dechesne at linaro.org>
> > > >> >>
> > > >>
> > > >>     wrote:
> > > >>      >
> > > >>      > >
> > > >>      > >
> > > >>      > >
> > > >>      > > On Wed, Jun 5, 2013 at 7:19 PM, Saul Wold <
> sgw at linux.intel.com
> > > >>     <mailto:sgw at linux.intel.com>> wrote:
> > > >>      > >
> > > >>      > >> On 06/05/2013 10:12 AM, Otavio Salvador wrote:
> > > >>      > >>
> > > >>      > >>>
> > > >>      > >>>
> > > >>      > >>>
> > > >>      > >>> On Wed, Jun 5, 2013 at 1:58 PM, Saul Wold
> > > >>     <sgw at linux.intel.com <mailto:sgw at linux.intel.com>
> > > >>      > >>> <mailto:sgw at linux.intel.com <mailto:sgw at linux.intel.com
> >>>
> > > >> wrote:
> > > >>      > >>>
> > > >>      > >>>     On 06/05/2013 09:32 AM, Nicolas Dechesne wrote:
> > > >>      > >>>
> > > >>      > >>>
> > > >>      > >>>         On Wed, Jun 5, 2013 at 6:30 PM, Saul Wold
> > > >>     <sgw at linux.intel.com <mailto:sgw at linux.intel.com>
> > > >>      > >>>         <mailto:sgw at linux.intel.com <mailto:
> > > >> sgw at linux.intel.com>>
> > > >>      > >>>         <mailto:sgw at linux.intel.com
> > > >>     <mailto:sgw at linux.intel.com> <mailto:sgw at linux.intel.com
> > > >>     <mailto:sgw at linux.intel.com>>>**>
> > > >>      > >>> wrote:
> > > >>      > >>>
> > > >>      > >>>              You could just directly put the
> nativesdk-libx11
> > > >>     in place
> > > >>      > >>>         of the
> > > >>      > >>>              variable, no need to have the variable
> there.
> > > >>      > >>>
> > > >>      > >>>
> > > >>      > >>>         yes, that's what I had initially, but found it
> was
> > > >>     less easy to
> > > >>      > >>>         read...
> > > >>      > >>>         with X11DEPENDS it's more 'obvious' that there is
> > > >>     something
> > > >>      > >>>         special..
> > > >>      > >>>         that said, i can make the change if that's really
> > > >> needed.
> > > >>      > >>>
> > > >>      > >>>     We do use the X11DEPENDS elsewhere when there are
> multiple
> > > >>      > >>>     dependencies, but I also found cases where we just
> > > >>     include the
> > > >>      > >>>     dependency directly in the test. I was trying pick a
> > > >>     direction:
> > > >>      > >>>     single entry no X11DEPENDS, multiple entries use
> > > >> X11DEPENDS.
> > > >>      > >>>
> > > >>      > >>>     Comments, flames, ...
> > > >>      > >>>
> > > >>      > >>>
> > > >>      > >>> Yes; I sent this patch in Febuary:
> > > >>      > >>>
> > > >>     http://patchwork.openembedded.****org/patch/44759/<http://**
> > > >> patchwork.openembedded.org/**patch/44759/<
> http://patchwork.openembedded.org/patch/44759/>
> > > >> >
> > > >>      > >>>
> > > >>      > >>> Please use this one instead of the recent one.
> > > >>      > >>>
> > > >>      > >>>
> > > >>      > >> Well reading back on that, it looks like I was waiting
> for an
> > > >>      > >> EXTRA_OECONF or related change to the autoconf scripts.
> > > >>      > >>
> > > >>      > >> Sau!
> > > >>      > >
> > > >>      > >
> > > >>      > > hmm. ok, sorry Otavio, i missed the other patch. I will
> check
> > > >>     on my side
> > > >>      > > too about EXTRA_OECONF.
> > > >>      > >
> > > >>      >
> > > >>      > Nicolas, don't worry. It is normal to end redoing some stuff.
> > > >>      >
> > > >>      > Last time I checked it had no support in Qt build system; I
> am
> > > >>     not sure if
> > > >>      > it uses or not the host headers (in case they exist) but it
> needs
> > > >>     testing
> > > >>      > to be sure.
> > > >>
> > > >>     I can confirm that it does not link against host X11 when built
> w/o
> > > >> that
> > > >>     dependency and nativesdk has no x11 libs/headers. As I
> previously
> > > >>     mentioned,
> > > >>     we've been using this fix for over 6 months on several releases
> built
> > > >> on
> > > >>     different machines w/o problems...
> > > >>
> > > >>
> > > >> In this case the patch can be merged 'as is'.
> > > >>
> > > >>
> > > > I am still concerned about a floating dependency here, imagine the
> > > > following:
> > > >
> > > > Build the toolchain with X11 enabled, nativesdk-libx11 is build, now
> > > > rebuild with X11 disabled, the dependency is gone, but the libraries
> and
> > > > headers still exist in the sysroot and thus the configure will still
> enable
> > > > x11 in qte, bad things happen.
> > > >
> > >
> > > I expect different distro features to have different build dirs. No?
> > >
> > >
> > > > We need to have a disable flag to autotools.
> > > >
> > >
> > > The qmake based system does not provide this; so to support it, it will
> > > need to be a hack ...
> >
> > I agree with Otavio here - seems like you are after a rare corner case
> here...
>
> We need determinism in the builds. Yes in this case it might be hard to
> achieve but I do think we need to achieve it.
>
> So this patch isn't going in until we can avoid this option magically
> floating, sorry. I've said this before and I've not changed my mind.
>

It is completely deterministic that it is *broken* for every one not using
a DISTRO with X11; this is deterministic enough for me to justify it go in.
As most of people didn't complain about this before, it is clear most use
*x11* feature and thus this won't affect them.

I see no point in delay it; complete determinism is better - sure - but a
broken Qt toolchain for x11less systems does not improve it.

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://projetos.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openembedded.org/pipermail/openembedded-core/attachments/20130606/9c540eed/attachment-0002.html>